Grammar correcting method and apparatus

Abstract

A grammar correcting method is provided, the method including receiving a sentence generated based on speech recognition, receiving information associated with a speech recognition result of the sentence and correcting grammar in the sentence based on the information associated with the speech recognition results of the sentence.

What is claimed is: 
     
       1. A processor-implemented method, performed by one or more processors of a speech recognition system, the method comprising:
 generating a speech-recognized sentence based on speech recognition of a speech; generating information associated with a speech recognition result of the speech-recognized sentence; 
 determining whether a grammatical error exists in the speech-recognized sentence, the grammatical error being associated with the speech recognition; 
 identifying a plurality of candidate words, associated with the determined grammatical error, from among a plurality of words included in the speech-recognized sentence, each of the plurality of candidate words being grammatically associated with each other; 
 in response to the grammatical error in the speech-recognized sentence being determined to be based on the plurality of candidate words, selecting a target word required to be corrected among the plurality of candidate words based on respective recognition reliabilities of the candidate words; and 
 generating a final speech-recognized sentence of the speech by correcting the selected target word to correct the grammatical error of the speech-recognized sentence based on at least one remaining candidate word excluding the target word, among the plurality of candidate words, 
 wherein the information associated with the speech recognition result of the speech-recognized sentence includes respective recognition reliabilities of each of words in the speech-recognized sentence, the respective recognition reliabilities being associated with the speech recognition, and 
 wherein the information associated with the speech recognition result of words included in the speech-recognized sentence is determined based on at least one of a recognition reliability associated with a pronunciation of the speech-recognized sentence and a recognition reliability associated with an intention of the speech-recognized sentence. 
 
     
     
       2. The method of  claim 1 , wherein the recognition reliabilities of the words in the speech-recognized sentence comprise recognition reliabilities of parts of the words. 
     
     
       3. The method of  claim 1 , wherein the information associated with the speech recognition result of the speech-recognized sentence is dependent on at least one of recognition reliabilities from a language model and recognition reliabilities from an acoustic model. 
     
     
       4. The method of  claim 1 , wherein the correcting the grammar in the speech-recognized sentence further comprises:
 detecting the plurality of candidate words associated with the grammatical error among the words in the speech-recognized sentence by verifying grammar in the speech-recognized sentence; 
 selecting at least one candidate word among the plurality of candidate words based on the respective recognition reliabilities of each of the words in the speech-recognized sentence included in the information associated with the speech recognition result of the speech-recognized sentence; and 
 correcting the at least one selected candidate word. 
 
     
     
       5. The method of  claim 4 , wherein the selecting the at least one candidate word comprises selecting a candidate word having a lowest degree of recognition reliability among the plurality of candidate words. 
     
     
       6. The method of  claim 1 , further comprising:
 generating the speech-recognized sentence by recognizing the speech based on a language model and an acoustic model in the speech recognition; and 
 providing the generated recognition reliabilities of the words in the speech-recognized sentence, for the correcting of the grammatical error in the speech-recognized sentence, dependent on respective recognition reliabilities from the language model and/or the acoustic model during the speech recognition. 
 
     
     
       7. The method of  claim 1 , wherein the speech-recognized sentence generated based on the speech recognition is a recognized sentence resulting from a decoder operation that considers results of an acoustic model and/or a language model. 
     
     
       8. The method of  claim 7 , wherein the recognition reliabilities of the words in the speech-recognized sentence are based on recognition reliabilities respectively determined by the acoustic model and/or the language model during the generation of the speech-recognized sentence in the speech recognition. 
     
     
       9. The method of  claim 8 , wherein
 the recognition reliabilities determined by the acoustic model are recognition reliabilities associated with the pronunciation of the speech-recognized sentence and the recognition reliabilities determined by the language model are recognition reliabilities associated with the intention of the speech-recognized sentence, and 
 the recognition reliabilities of the words include the recognition reliabilities associated with the pronunciation of the speech-recognized sentence and the recognition reliabilities associated with the intention of the speech-recognized sentence. 
 
     
     
       10. A computer program combined with hardware and stored in a non-transitory computer-readable storage medium to cause the one or more processors of the speech recognition system to perform the method of  claim 1 . 
     
     
       11. A speech recognition system, the system comprising
 one or more processors configured to:
 verify grammar in a speech-recognized sentence, generated based on speech recognition of a speech, for grammatical error correction; 
 determine whether grammatical errors of the sentence exist based on a plurality of candidate words, each of the plurality of candidate words being identified in the verification from words in the sentence and being grammatically associated with each other, and the grammatical errors being associated with the speech recognition; 
 in response to the grammatical errors in the speech-recognized sentence being determined to be based on the plurality of candidate words, select a target word required to be corrected among the plurality of candidate words based on respective recognition reliabilities of the candidate words, the respective recognition reliabilities being associated with the speech recognition; and 
 generate a final speech-recognized sentence of the speech by correcting the selected target word to correct the grammatical error of the sentence based on at least one remaining candidate word excluding the target word, among the plurality of candidate words, 
 wherein the information associated with the speech recognition result of words included in the speech-recognized sentence is determined based on at least one of a recognition reliability associated with a pronunciation of the speech-recognized sentence and a recognition reliability associated with an intention of the speech-recognized sentence. 
 
 
     
     
       12. The system of  claim 11 , wherein, for the verification, the one or more processors are further configured to detect the plurality of candidate words associated the grammatical errors among the words comprised in the speech-recognized sentence. 
     
     
       13. The system of  claim 11 , wherein the one or more processors are further configured to select at least one candidate word among the plurality of candidate words associated with the grammatical errors based on the recognition reliabilities of the words in the speech-recognized sentence, and correct the at least one selected candidate word. 
     
     
       14. The system of  claim 13 , wherein the one or more processors are further configured to select a candidate word having a lowest degree of recognition reliability among the plurality of candidate words. 
     
     
       15. The system of  claim 11 , wherein the recognition reliabilities of the words in the speech-recognized sentence are dependent on at least one of recognition reliabilities from a language model and recognition reliabilities from an acoustic model. 
     
     
       16. The system of  claim 11 , wherein
 the one or more processors are further configured to:
 perform a speech recognition operation of the speech based on a language model and an acoustic model to generate the speech-recognized sentence; and 
 provide the recognition reliabilities of the words in the speech-recognized sentence, for the correcting of the grammatical errors, dependent on respective recognition reliabilities from the language model and/or the acoustic model during the speech recognition operation. 
 
 
     
     
       17. The system of  claim 11 , wherein the speech-recognized sentence is a generated result of a speech recognition operation, with the generated result of the speech recognition operation being a recognized sentence resulting from a decoder operation that considers results of an acoustic model and/or a language model. 
     
     
       18. The system of  claim 17 , wherein the recognition reliabilities of the words in the speech-recognized sentence are based on recognition reliabilities respectively determined by the acoustic model and/or the language model during the generation of the speech-recognized sentence in the speech recognition operation. 
     
     
       19. The system of  claim 18 , wherein
 the recognition reliabilities determined by the acoustic model are recognition reliabilities associated with the pronunciation of the speech-recognized sentence and the recognition reliabilities determined by the language model are recognition reliabilities associated with the intention of the speech-recognized sentence, and 
 the recognition reliabilities of the words include the recognition reliabilities associated with the pronunciation of the speech-recognized sentence and the recognition reliabilities associated with the intention of the speech-recognized sentence. 
 
     
     
       20. A speech recognition system, the system comprising:
 one or more processors configured to:
 generate a speech-recognized sentence based on speech recognition of a speech; generate information associated with a speech recognition result of the speech-recognized sentence; 
 determine whether a grammatical error exists in the speech-recognized sentence, the grammatical error being associated with the speech recognition; 
 identify a plurality of candidate words, associated with the determined grammatical error, from among a plurality of words included in the speech-recognized sentence, each of the plurality of candidate words being grammatically associated with each other; 
 in response to the grammatical error in the speech-recognized sentence being determined to be based on the plurality of candidate words, select a target word required to be corrected among the plurality of candidate words based on respective recognition reliabilities of the candidate words; 
 generating a final speech-recognized sentence of the speech by correcting the selected target word to correct the grammatical error of the speech-recognized sentence based on at least one remaining candidate word excluding the target word, among the plurality of candidate words, 
 
 wherein the information associated with the speech recognition result of the speech-recognized sentence includes respective recognition reliabilities of each of the words in the speech-recognized sentence, the respective recognition reliabilities being associated with the speech recognition, and 
 wherein the information associated with the speech recognition result of words included in the speech-recognized sentence is determined based on at least one of a recognition reliability associated with a pronunciation of the speech-recognized sentence and a recognition reliability associated with an intention of the speech-recognized sentence. 
 
     
     
       21. The system of  claim 20 , wherein the one or more processors are further configured to:
 detect the plurality of candidate words associated with the grammatical error included in the speech-recognized sentence by verifying grammar in the speech-recognized sentence; 
 select at least one candidate word among the plurality of candidate words based on the respective recognition reliabilities of each of the words in the speech-recognized sentence included in the information associated with the speech recognition result; and 
 correct the at least one selected candidate word. 
 
     
     
       22. The system of  claim 20 , wherein the speech-recognized sentence generated based on the speech recognition is a recognized sentence resulting from a decoder operation that considers results of an acoustic model and/or a language model. 
     
     
       23. The system of  claim 22 , wherein the recognition reliabilities of the words in the speech-recognized sentence are based on recognition reliabilities respectively determined by the acoustic model and/or the language model during the generation of the speech-recognized sentence in the speech recognition. 
     
     
       24. The system of  claim 23 , wherein
 the recognition reliabilities determined by the acoustic model are recognition reliabilities associated with the pronunciation of the speech-recognized sentence and the recognition reliabilities determined by the language model are recognition reliabilities associated with the intention of the speech-recognized sentence, and 
 the recognition reliabilities of the words include the recognition reliabilities associated with the pronunciation of the speech-recognized sentence and the recognition reliabilities associated with the intention of the speech-recognized sentence.